Platform · Insights

Insights from every visit

Insights shows what visitors actually do across your whole website: which pages work, which hold people back, and why.

Your whole website, understood

Know what works on your website, and what doesn’t

One clear picture of every page, not a dashboard to dig through.

How every page performs, where visitors convert and where they get stuck, read from friction, errors, scroll, attention, speed, paths and funnels, with an AI Analyst that shows its work.

  • One overview of the whole site, with only real changes highlighted
  • A health check for every kind of page, worst first
  • Journeys and funnels from every visit, with no tagging plan for page steps
  • Opportunities ranked from behaviour, customer feedback and past tests
  • An AI Analyst whose every figure links to its source
All plansAbout 5 KBNo cookies
The Insights overview for one website over 28 days: a tile per figure, sessions, conversion rate, visits with an error, dead clicks, rage clicks, scroll past halfway and approximate web vitals, with only significant changes coloured, and the three biggest changes ranked.

Pages

Which page is hurting you, and how

Every kind of page, from product page to checkout, or every URL template, gets a status against the site average: errors, friction, exits above average, slow, healthy or not enough data. The worst sort first, and each opens into a full page detail.

  • Rage clicks, errors, exit rate, approximate LCP and engaged time against the site average
  • Attention by depth: where on the page engaged time is spent
  • The most seen elements, and whether converting differs for visitors who saw them
  • Customer comments, past experiments and open opportunities for that page in one place
The Pages screen: each kind of page with a status against the site average, a product page flagged for friction, checkout for errors, category for exits above average and pricing as slow, with home and sign-up healthy and the blog without enough data.

Journeys and funnels

Ask a new question about last month

Funnels are defined after the fact, from data already collected, so a step you never tagged can still be measured. Journeys show the paths visitors actually take, and any path becomes a funnel in one click.

  • Steps by page type, URL pattern or any event the site already sends
  • Split by device, visitor type, traffic source, campaign and language
  • A saved funnel gets 120 days of history built in the background
  • Paths and funnels are built overnight for each complete day
A saved sign-up funnel from pricing to account created, with visits and conversion at each step, the biggest drop at the company details step, and the previous period alongside.

Funnel diagnostics

What is different about the visitors who leave

Open a step and compare the visits that leave with those that carry on, on device, entry page, errors, dead clicks, slow loading, scroll, campaign and more. Each difference is tested and corrected for multiple comparisons, and each is something to test, not a proven cause.

  • Ranked differences between visitors who leave and visitors who continue
  • Where leavers went instead, and how long the step took
  • Whether people who leave come back and finish within 7 or 30 days
  • Draft experiment from any finding, with the hypothesis filled in
Funnel diagnostics for the company details step: the visits that leave compared with those that continue, ranked, with dead clicks on the company size field and mobile traffic standing out, one difference marked as not significant, and a button to draft an experiment.

Opportunities

A ranked list of what to work on next

Pertento finds the pages where visitors hit far more friction or errors than elsewhere, and the pages drawing unusually negative survey comments, then ranks them by reach, severity and how many kinds of evidence agree. Code finds them with a statistical test; the AI only writes the wording.

  • Friction hotspots, with the most affected elements named
  • Error hotspots, with the top JavaScript errors named
  • Pages drawing an unusual share of negative survey comments
  • The evidence behind each: observed, customers report, tested
The Opportunities list: a friction hotspot on mobile product pages, with the size selector named, its reach, and evidence observed and reported by customers but not yet tested, then an error hotspot on checkout and a page customers complain about.

Analyst

An AI analyst that shows its work

Ask in plain language and get an answer whose every figure links to the screen it came from. The Analyst reads the same data as the screens through read-only tools, and a number that is not in the data is removed before the answer is shown.

  • Suggested questions built from the site’s current state
  • Small tables and charts in the answer, and follow-ups
  • Conversations shared with the team, and an experiment drafted from any answer
  • An answer based on a sample says so, with its margin of error
An Analyst conversation: a question about friction on the mobile product page, answered with figures that each link to the screen they came from, a note that the difference is worth testing rather than a proven cause, and an option to draft an experiment.

Running experiments

Catch the variant that makes the page worse

Diagnose compares each variant with control on errors, dead and rage clicks, scroll reach and approximate web vitals, so a result can be explained. The behaviour guardrail watches every running experiment and warns when a variant makes the site worse for visitors.

  • Each variant against control, by device or page type
  • The top errors and most dead-clicked elements per variant
  • Checked every hour; a clearly broken variant is typically flagged within about a day
  • It warns, and never stops or changes a test by itself
Diagnose on a running experiment: each behaviour figure for control against the variant, with the variant showing far more dead clicks and slower loading, and a guardrail warning that leaves the test running.

Statistics you can trust

Noise is not coloured as a win

Insights treats behaviour with the same care as the experiment statistics. These are the rules behind every highlighted change.

01Tested, then highlightedA change is coloured green or red only when it is significant, tested with a two-proportion test at 95%. Everything that could be noise stays neutral.
02Corrected for many checks at onceComparing many metrics or segments at the same time makes a few look different by chance alone, so differences are corrected for multiple comparisons with Benjamini-Hochberg.
03“Not enough data”, said plainlyA rate needs at least 200 visits in each period, and a speed figure 1,000 samples, before a change is called real. Below that, the screen says there is not enough data rather than guessing.
04Differences, not causesWhat separates the visitors who leave from those who stay is shown as a difference worth testing, never as the proven cause of a drop.
05Sampling disclosedA read over more than 500,000 visits uses a sample and says so, with its margin of error. The overview, Pages, Journeys and saved funnels are never sampled.
06Approximate vitals, labelled soLCP, INP and CLS are measured by the snippet and labelled approximate. They will not match Google’s field data exactly, and the product never pretends they do.

Light on the page

Careful with the page and the data

Behaviour is collected by a small add-on to the Pertento script, and only on websites where behaviour analysis is switched on.

01About 5 KBBehaviour is collected by an add-on to the Pertento script of about 5 KB compressed, capped there by an automated size check. Websites without behaviour analysis never download it.
02Loads after the page is shownIt loads asynchronously once the page is visible, so it never delays rendering or experiment variants, and its heavier work runs in idle time.
03No cookiesIt sets no cookies and writes nothing to localStorage or sessionStorage.
04No form values, no page textIt never reads what visitors type or the text on the page. The only text it reads is a button or link label, and numbers, email addresses and URLs are scrubbed from it.
05Pages as patternsFull URLs and query strings are never stored. Pages are kept as patterns such as /products/:id, and an arrival keeps only the referring domain and the UTM source, medium and campaign.
06Your consent setup appliesCollection follows whatever consent setup the website already applies to the Pertento script.

Questions

Questions about Insights

Is this session replay?

No. Insights does not record or replay visits and does not draw heatmaps. It measures behaviour as figures, friction, errors, scroll, attention, speed, paths and funnels, and turns them into opportunities to test.

How is it different from our analytics tool?

An analytics tool reports traffic and conversions. Insights looks for where visitors struggle, ranks it into opportunities with the evidence behind each, and drafts the experiment, in the same product that runs the test and reads its result.

Will it slow our website down?

It is built not to. The add-on is about 5 KB, loads after the page is shown and does its heavier work in idle time. In lab tests, interaction responsiveness with it on was not different from off beyond run-to-run noise.

Does it tell us why visitors drop out?

It shows what is different about the visitors who leave a step, tested and corrected for multiple comparisons, and the Analyst can talk you through it. Those are differences worth testing rather than proven causes, which is why every finding can become an experiment.

How fresh is the data?

The overview, Pages and the friction figures update within minutes. Journeys, funnels, time on page and conversion are built overnight for each complete day, and daily figures go back 13 months.

Which plans include it?

All of them. Behaviour analysis is switched on per website. The AI features, such as the Analyst, use the account’s AI credits as they do everywhere else in Pertento.

How is it different from Learnings?

Insights reads what visitors do on your website. Learnings reads what your finished experiments taught. Insights shows where to look and what to test; Learnings keeps what each test showed and suggests the next one.

Everything in the box

The full feature set

Experimentation

Statistics

Monitoring

Voice of customer

Learnings

Insights

Pertento is a conversion-rate-optimisation platform for websites and online stores, built for both in-house teams and CRO agencies running experimentation programmes across a client roster.

Integrations

Works with the stack you already run

One snippet drops into any site, or go fully server-side. Nothing to re-architect.

01 / Snippet

One-line snippet

Paste one tag in your head and you are live: a 0.9 KB loader, no build step, no dependencies.

Paste it before </head>
02 / Tag manager

Google Tag Manager

Deploy and manage experiments straight through GTM. No developer time required.

Add it as a Custom HTML tag
03 / Server

Server-side API

Run tests beyond the browser: pricing, search and routing, flicker-free.

Install the SDK
  • Shopify
  • BigCommerce
  • Shopware
  • Centra
  • Geins
  • Saleor
  • WordPress
  • Optimizely CMS
  • Storyblok
  • Next.js
  • Vue
  • Astro
  • Matomo
  • Amplitude
  • RudderStack
  • Slack
  • WooCommerce
  • Salesforce
  • Wix
  • Litium
  • commercetools
  • Medusa
  • Drupal
  • Contentful
  • Webflow
  • Nuxt
  • Angular
  • Remix
  • Piwik PRO
  • Mixpanel
  • Tag Manager
  • Webhooks
  • Adobe Commerce
  • PrestaShop
  • Squarespace
  • Norce
  • Shopify Hydrogen
  • Vendure
  • Umbraco
  • Sanity
  • Framer
  • React
  • SvelteKit
  • GA4
  • Adobe Analytics
  • Segment
  • Klaviyo

Not on the list? If it renders HTML, Pertento can test it: read how each one connects ↗

Better experiments, better conversions

Test on all visitors with the world’s lightest script and make confident decisions powered by real-time reporting.Start trialRequest a demo